Towards Transabdominal Fetal Pulse Oximetry: Computational Modeling and Algorithm Development

Non-invasive fetal oxygen monitoring is critical for assessing fetal well-being and preventing adverse outcomes caused by fetal hypoxia. Current clinical standards, such as cardiotocography (CTG), primarily rely on measurements of fetal heart rate and maternal contractions, both of which are indirect indicators of actual fetal oxygenation status. As a result, CTG often produces false-positive results, leading to unnecessary medical interventions. Transabdominal fetal pulse oximetry, in contrast, offers the potential for direct, continuous, and non-invasive estimation of fetal oxygen saturation (SpO2). However, significant physiological and technical challenges—including strong maternal tissue interference, complex multi-layer photon propagation, and motion artifacts—have limited its clinical adoption. This dissertation addresses these challenges and advances transabdominal fetal pulse oximetry toward clinical feasibility through computational modeling and algorithmic innovations. First, a novel self-calibrated pulse oximetry algorithm was developed to overcome a fundamental limitation: conventional pulse oximeters re quire empirical calibration on healthy subjects, resulting in inaccuracies when applied to low-oxygen saturation conditions like those in fetal circulation. The proposed algorithm significantly improves accuracy in the low SpO2 range relevant for fetal applications. Next, the self-calibrated algorithm was extended by incorporating multi layer photon transport modeling, which allows accurate fetal SpO₂ estimation despite significant maternal tissue interference. To enhance real-world applicability, extensive sensitivity analyses based on MRI-derived anatomical models were conducted to optimize device configuration and evaluate the effects of fetal positioning. Finally, a deep learning framework utilizing long short-term memory networks was developed to robustly remove motion artifacts from pulsatile optical signals, further improving measurement reliability in clinical environments. Together, these advancements provide a solid foundation for accurate, continuous, and non-invasive fetal SpO₂ monitoring, creating pathways toward safer and more effective fetal health monitoring in clinical practice. Additionally, these methodologies offer valuable tools applicable to a wide range of other optical physiological monitoring applications.